Social Networks (IT-703 (C)) - Unit 2 Short Notes

UNIT 2: SOCIAL NETWORKS - EXAM-FOCUSED SHORT NOTES

I. FOUNDATIONS OF THE SOCIAL WEB AND SOCIAL NETWORKS

A. Emergence of the Social Web

[!TIP] Exam Focus: Be prepared to contrast Web 1.0 vs. 2.0 and list key characteristics.

The Social Web represents the transition from static, read-only Web 1.0 (static HTML pages, company-controlled content) to interactive, user-generated Web 2.0 and beyond.

• Key Milestones & Driving Forces:

  • Broadband普及 & Mobile Internet: Enabled constant connectivity.

  • Platform Evolution: From forums (1990s) to Social Networking Sites (SNS) like Friendster, MySpace, then Facebook/LinkedIn (mid-2000s), to ubiquitous apps (2010s+).

  • Technology: AJAX (asynchronous updates), APIs for integration, cloud computing for scalability.

• Core Characteristics:

  • User-Generated Content (UGC): Users create and share text, images, video.

  • Participation & Collaboration: Wikis, crowdsourcing, social curation.

  • Network Effects: Value increases as more users join.

  • Rich Interactions: Likes, shares, comments, @mentions.

B. Importance and Scope of Social Network Analysis (SNA)

[!TIP] Common Pitfall: Don't just define SNA; always link it to why it's useful for a given domain. Social Network Analysis (SNA) is the process of investigating social structures through the use of network and graph theory. It characterizes networks in terms of nodes (actors) and ties/edges (relationships).

• Purpose: To understand patterns of relationships, identify influential actors, map information flow, and detect subgroups/communities.

• Applications:

  • Sociology: Studying diffusion of innovations, social capital.

  • Business: Marketing (influencer identification), organizational analysis (communication bottlenecks).

  • Health: Tracking disease spread, modeling patient support networks.

  • Security: Identifying terrorist cells, fraud rings in financial networks.

  • Computer Science: Recommender systems, web structure analysis.

C. Types of Web-Based Networks

[!TIP] Exam Tip: Be able to give a concrete example for each type.

Networks can be classified by the nature of their nodes and edges:

II. SEMANTIC WEB AND ONTOLOGIES FOR SOCIAL DATA

A. Resource Description Framework (RDF) and RDF Schema

[!TIP] Key Formula: RDF is built on the Triple: (Subject, Predicate, Object).

• RDF: A standard for data interchange on the web. It represents information as directed, labeled graphs.

  • Triple: The fundamental unit. Example: (Alice, knows, Bob).

  • Graph Model: A set of triples forms a graph where subjects and objects are nodes, predicates are directed edges.

• RDF Schema (RDFS): A vocabulary description language that provides a basic type system for RDF.

  • Defines classes (e.g., foaf:Person) and properties (e.g., foaf:knows).

  • Allows subClassOf and subPropertyOf hierarchies.

  • Role in Social Data: Provides a formal way to define social ontologies (like FOAF) so that data from different sources can be integrated and understood by machines.

B. Web Ontology Language (OWL) - Unique Features

[!TIP] Distinguish: RDFS is for simple vocabularies; OWL is for rich, computable knowledge. OWL is a knowledge representation language for authoring ontologies on the Semantic Web. It builds on RDF/RDFS but is more expressive.

• Unique Features:

  1. Greater Expressiveness: Can define complex class relationships (e.g., ClassA ≡ ClassB ⊓ hasSome Property Value).

  2. Formal Semantics: Has a precise mathematical model-theoretic semantics, enabling logical reasoning.

  3. Reasoning Support: Allows automated inference (e.g., if A isParentOf B and B isParentOf C, infer A isGrandparentOf C).

• OWL Dialects:

  • OWL Lite: For simple taxonomies and constraints. Easier to implement.

  • OWL DL (Description Logic): Maximum expressiveness while retaining computational completeness and decidability. Most widely used.

  • OWL Full: Maximum expressiveness, no computational guarantees. Compatible with RDF fully.

C. FOAF (Friend of a Friend) - Ontological Representation

[!TIP] Remember: FOAF is an RDF vocabulary, not a full OWL ontology. FOAF is a machine-readable ontology that describes people, their activities, and their relations to other people/things.

• Core Classes: foaf:Person, foaf:Organization, foaf:Group.

• Key Properties for Social Relationships:

  • foaf:knows: A social acquaintance relationship (asymmetric, e.g., Alice knows Bob, but Bob may not know Alice).

  • foaf:based_near: Approximate geographic location.

  • foaf:member: Links a person to a group.

  • foaf:workplaceHomepage, foaf:mbox (email).

• Role: Provides a decentralized standard for social profile data. A user's FOAF file (hosted on their own server) can list their friends by their FOAF URIs, enabling a web of decentralized social networks that can interoperate, unlike walled gardens like Facebook.

III. COMMUNITY DETECTION AND EVOLUTION IN SOCIAL NETWORKS

A. Conceptual Definitions of Community

[!TIP] Crucial Distinction: Local definitions find sub-structures; global definitions partition the whole graph.

A community (or module/cluster) is a subset of nodes with dense connections internally and sparser connections externally.

• Local Community Definitions:

  • Clique: A subset where every node is connected to every other node (maximal density).

  • k-core: A maximal subgraph where each node has degree ≥ k within the subgraph. (e.g., 3-core).

  • k-clan/k-plex: Relaxations of cliques allowing some missing ties.

• Global Community Definitions:

  • Based on optimizing a quality function like Modularity (Q). Seeks a partition of the entire network that maximizes intra-community edges vs. expected random distribution.

• Vertex-Centric Definitions:

  • Ego Network: The network consisting of a focal node ("ego") and its immediate neighbors ("alters"), plus the ties among alters.

  • Structural Equivalence: Nodes that have identical connection patterns to the rest of the network (even if not directly connected).

B. Measuring Evolution of Web Communities

[!TIP] Key Metrics: Track changes in size, density, membership over sequential time snapshots.

When analyzing a series of network snapshots (e.g., from web archives):

1. Extract Temporal Community Structure: Apply a community detection algorithm (e.g., modularity-based) to each snapshot.

2. Evolution Metrics:

   • Size: Number of nodes/edges in a community over time.

   • Density: Internal edge density of the community.

   • Membership Turnover: Jaccard similarity of node sets between consecutive snapshots. Low similarity = high churn.

   • Growth/Decay/Transformation: Track if communities expand, shrink, split, merge, or dissolve.

   • Persistence: How long a community (by label or core membership) survives.

C. Network Reduction Techniques

[!TIP] Goal: Reduce complexity while preserving the property of interest (e.g., community structure).

• Purpose: To simplify large, dense networks for visualization, faster computation, or to focus on a specific structural aspect.

• Common Methods:

  • Ego Network Extraction: Isolates the network around a specific node (ego) and its direct neighbors. Preserves local structure.

  • k-core Decomposition: Iteratively removes nodes with degree < k. The remaining k-core is the maximal subgraph with minimum degree k. Prunes peripheral nodes, revealing the dense core.

  • Backbone Extraction: Removes "noise" edges based on weight thresholds (e.g., in weighted networks, keep only edges above a certain strength percentile). Methods like Disparity Filter.

  • Community-Aggregation: Collapse each detected community into a single "super-node" and represent inter-community ties as edges between super-nodes. Creates a meta-network.

IV. SOCIAL NETWORK ANALYSIS: METHODS AND REPRESENTATIONS

A. Centrality Measures

[!TIP] Mnemonic: Degree (popularity), Betweenness (brokerage), Closeness (reach), Eigenvector (influence).

Quantify the "importance" or "prominence" of a node.

B. Clustering and Community Detection

• Clustering Coefficient:

  • Local (for node v): $$\displaystyle C(v) = \frac{2T(v)}{deg(v)(deg(v)-1)} $$ <br> Where $T(v)$ is number of triangles through v. Measures "cliquishness" of immediate neighborhood.

  • Global (Transitivity): $$\displaystyle C = \frac{3 \times \text{number of triangles}}{\text{number of connected triples}} $$. Overall network tendency to form clusters.

• Community Detection Algorithms:

  • Modularity Optimization: (e.g., Louvain algorithm). Maximizes $$\displaystyle Q = \frac{1}{2m} \sum_{ij} \left[ A_{ij} - \frac{k_i k_j}{2m} \right] \delta(c_i, c_j) $$, where $$\displaystyle A_{ij} $$ is adjacency, $$\displaystyle k_i $$ degree, $m$ total edges, $\delta$=1 if same community.

  • Hierarchical Clustering: Agglomerative (merge nodes) or divisive (split network) based on edge betweenness.

  • Overlapping vs. Non-overlapping: Some algorithms (e.g., Clique Percolation) allow a node to belong to multiple communities, reflecting real-world multi-group memberships.

C. Matrix Representations of Networks

[!TIP] Essential Tool: Matrices enable spectral analysis (eigenvalues/vectors) for SNA metrics.

• Adjacency Matrix ($A$):

  • $$\displaystyle A_{ij} = 1 $$ if edge from $i$ to $j$, else $0$.

  • Symmetric for undirected graphs, asymmetric for directed.

  • Can be weighted ($$\displaystyle A_{ij} = w_{ij} $$).

• Incidence Matrix ($B$):

  • Rows = nodes, Columns = edges.

  • $$\displaystyle B_{ve} = 1 $$ if node $v$ is incident to edge $e$, else $0$. (For undirected, often uses +1/-1 for orientation).

• Laplacian Matrix ($L$):

  • $$\displaystyle L = D - A $$, where $D$ is the diagonal degree matrix ($$\displaystyle D_{ii} = deg(i) $$).

  • Properties: Symmetric, positive semi-definite, row/column sums = 0. Eigenvalues reveal structural properties (e.g., number of connected components = multiplicity of eigenvalue 0).

  • Applications: Spectral clustering (using eigenvectors of $L$), graph partitioning, random walks.

V. APPLICATIONS AND ENABLING TECHNOLOGIES IN SOCIAL CONTEXTS

A. Social Networks as Platforms

• Facilitation: Enable asynchronous/synchronous interaction (posts, chat), information sharing (feeds), and collaboration (groups, events).

• Network Effects & Virality: Direct (more users → more value) and indirect (complementary goods) effects. Virality is the rapid, exponential spread of content, modeled by epidemic thresholds and cascade models (e.g., Independent Cascade, Linear Threshold).

• Role: Fundamental infrastructure for modern communication, news dissemination, political mobilization, and economic activity (creator economies, social commerce).

B. Reality Mining

[!TIP] Definition: The collection and analysis of real-world behavioral data from sensors (primarily mobile phones).

• Techniques: Bluetooth proximity (detecting co-located people), GPS/ Cell tower location, call/SMS logs, app usage, social media check-ins.

• Applications:

  • Urban Planning: Understanding human mobility patterns, traffic flow.

  • Epidemiology: Modeling disease spread based on contact networks.

  • Behavioral Studies: Quantifying social interaction patterns, stress levels (via voice analysis).

C. Context Awareness

• Incorporates: Location, Time, Activity, Social Context (who you are with), Device state.

• Enables: Adaptive services and personalized experiences (e.g., context-aware reminders, location-based recommendations, automatically setting phone to silent in a meeting).

• Technologies Stack:

  1. Sensors: GPS, accelerometer, microphone, camera.

  2. Context Models & Ontologies: Formal representation (e.g., using OWL) of context concepts and relationships.

  3. Reasoning & Rule-Based Systems: Infer high-level context (e.g., "in a meeting") from low-level sensor data using rules or machine learning.

VI. PRIVACY, SECURITY, AND ATTACKS IN ONLINE SOCIAL NETWORKS (OSNs)

A. Privacy Issues in OSNs

• Data Collection & Profiling: Platforms collect vast amounts of explicit (profile) and implicit (behavioral, location) data to build detailed user profiles for advertising.

• Disclosure Risks:

  • Personal: Home address, phone number, birthdate.

  • Sensitive: Health conditions, political/religious views, financial status.

  • Re-identification: Anonymized data can often be re-linked to individuals using auxiliary information.

• Settings & Control Challenges: Complex, frequently changing privacy settings; defaults are often public; social conundrum (hiding info from friends vs. platform).

B. Attack Spectrum (in OSNs)

[!TIP] Memorize the 5 P's: Plain, Profile Cloning, Profile Hijacking, Porting, Censorship.

1. Plain Impersonation: Creating a fake profile with fabricated identity details (e.g., fake celebrity account).

2. Profile Cloning: Copying a legitimate user's public profile information (name, photo, friends list) to create a fraudulent duplicate profile for scams or slander.

3. Profile Hijacking: Gaining unauthorized access to a real user's account (via credential theft, session hijacking) and taking control of it.

4. Profile Porting: Transferring a user's social identity and connections from one platform to another (often without consent), disrupting network integrity.

5. Censorship Attacks: Maliciously suppressing content or manipulating visibility (e.g., mass-reporting to trigger automated takedowns, Sybil attacks to downvote/drown out specific content).

C. Countermeasures Against Attacks

• Authentication: Multi-Factor Authentication (MFA), behavioral biometrics (typing rhythm, touch dynamics).

• Profile Integrity: Verification badges (for public figures), duplicate profile detection algorithms (comparing profile images, friend networks, activity patterns).

• Anomaly Detection: Monitoring for unusual login locations, sudden changes in posting behavior, or rapid friend-adding.

• Platform Policies & User Education: Clear terms of service, reporting mechanisms, user awareness campaigns about phishing and privacy settings.

D. Challenges for Decentralized OSNs (DOSNs)

[!TIP] Core Tension: Decentralization (privacy, control) vs. Centralized convenience (moderation, UX).

• Architectural: No central authority for global moderation, content removal, or coordinated security responses.

• Privacy Preservation: Harder to enforce consistent privacy policies across independent servers (pods). Data may be replicated across the network.

• Interoperability & Identity: Ensuring different DOSN software instances (e.g., Mastodon, PeerTube) can communicate (often via ActivityPub protocol). Decentralized identity (e.g., DID) is complex.

• Resistance to Censorship & Attacks: While resistant to platform-level censorship, vulnerable to distributed attacks (e.g., spam across many pods) and lack of centralized takedown for illegal content.

VII. SUPPORTING TOOLS AND PROTOCOLS

A. Email Groups as Social Networks

• Structure: A mailing list is a single email address that distributes messages to a list of subscribers. The network can be constructed where:

  • Nodes: Email addresses (users).

  • Edges: Directed from sender to recipient(s) for each email in a thread.

• Analysis: Reveals core-periphery structures (active posters vs. lurkers), community formation around topics, leadership (who initiates threads), and information flow patterns.

• Historical Significance: One of the earliest forms of large-scale online social interaction (1970s-80s, e.g., ARPANET lists). Modern equivalents: Slack/Discord channels, but email's asynchronous, broadcast nature is unique.

B. RSS Feeds

[!TIP] Definition: Really Simple Syndication (or Rich Site Summary).

• Purpose: A web feed format allowing users and applications to access updates to online content (blogs, news sites, podcasts) in a standardized, machine-readable XML format.

• Role in Social Context:

  • Information Dissemination: Enables users to "subscribe" to content sources, aggregating updates in a feed reader (e.g., Feedly).

  • Community Building: Bloggers and content creators use RSS to syndicate their work, building an audience that follows them across platforms.

  • Pushes vs. Pulls: Shifts from user actively checking sites (pull) to content being delivered to their aggregator (push), a precursor to social media news feeds.

C. Other Relevant Protocols/Tools (Brief)

• ActivityPub: The key federated protocol underlying DOSNs like Mastodon (microblogging) and PeerTube (video). Allows different server instances to communicate and share content/users.

• Webmention: A simple protocol for web-level interactions (likes, shares, comments) across different websites, enabling a decentralized "social web" without a central platform.

• Atom: An alternative syndication format to RSS, with similar use cases.

Final Exam Strategy: For 7-mark questions, structure your answer: 1. Clear Definition, 2. Key Components/Mechanism, 3. Example/Application, 4. Significance/Limitation. Always connect theory to the social networking context.